Content uploaded by Rodney Thomas Fernandez
Author content
All content in this area was uploaded by Rodney Thomas Fernandez
Content may be subject to copyright.
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
Green Roof Stormwater Retention: Effects of Roof Surface, Slope, and Media Depth
Nicholaus D. VanWoert, D. Bradley Rowe,* Jeffrey A. Andresen, Clayton L. Rugh,
R. Thomas Fernandez, and Lan Xiao
ABSTRACT
thus saving on energy consumption (Niachou et al., 2001;
Wong et al., 2003); increase the life span of a typical roof
Urban areas generate considerably more stormwater runoff than
by protecting the roof components from damaging ultra-
natural areas of the same size due to a greater percentage of impervi-
ous surfaces that impede water infiltration. Roof surfaces account for
violet rays, extreme temperatures, and rapid tempera-
a large portion of this impervious cover. Establishing vegetation on roof-
ture fluctuations (Giesel, 2001); filter harmful air pol-
tops, known as green roofs, is one method of recovering lost green
lutants (Liesecke and Borgwardt, 1997); provide a more
space that can aid in mitigating stormwater runoff. Two studies were
aesthetically pleasing environment to live and wo rk; pro-
performed using several roof platforms to quantify the effects of
vide habitat fo r a ran ge of o rgani sms from microbes to
various treatments on stormwater retention. The first study used three
birds (Brenneisen, 2 003; Gedge, 2003); and h ave th e po-
different roof surface treatments to quantify differences in stormwater
tential to reduce the Urban Heat Island Effect (Dimoudi
retention of a standard commercial roof with gravel ballast, an exten-
and Nikolopoulou, 2003; Rosenfeld et al., 1998; Wong
sive green roof system without vegetation, and a typical extensive
et al., 2003).
green roof with vegetation. Overall, mean percent rainfall retention
ranged from 48.7% (gravel) to 82.8% (vegetated). The second study
However, many consider stormwater runoff mitiga-
tested the influence of roof slope (2 and 6.5%) and green roof media
tion to be the primary benefit of green roofs due to the
depth (2.5, 4.0, and 6.0 cm) on stormwater retention. For all combined
prevalence of impervious surfaces in u rban and com mer-
rain events, platforms at 2% slope with a 4-cm media depth had the
cial areas and a failing stormwater management infra-
greatest mean retention, 87%, although the difference from the other
structure (Liptan, 2003). Rapid runoff from roofs and
treatments was minimal. The combination of reduced slope and deeper
other impervious surfaces can exacerbate flooding, in-
media clearly reduced the total quantity of runoff. For both studies,
crease erosion, and result in combined sewer overflows
vegetated green roof systems not only reduced the amount of storm-
that could potentially discharge raw sewage directly into
water runoff, they also extended its duration over a period of time
our waterways. Green roofs help mitigate the impact of
beyond the actual rain event.
high-density commercial and residential development
by restoring displaced vegetation. Studies have shown
that green roofs can absorb water and release it slowly
U
rban stormwater runoff has come to the fore-
over a period of time as opposed to a conventional roof
front as an environmental concern. The USEPA
where stormwater is immediately discharged (Liesecke,
has indicated that a typical city block generates more
1999; Moran et al., 2003; Schade, 2000). Research has
than five times as much runoff than a woodlot of the
indicated that an extensive green roof, depending on sub-
same size (USEPA, 2003). Urban stormwater runoff car-
strate de pth, c an retain 60 to 100% of incoming rainfall
ries with it numerous environmental contaminants in-
(Liesecke, 1998; Monterusso et al., 2004; Schade, 2000).
cluding pesticides, heavy metals, and nutrients, which
This reduction in quantity of runoff from a green roof
may eventually flow into lakes and streams (Bucheli
leads to improved stormwater runoff and surface water
et al., 1998; Mason et al., 1999). According to the USEPA
quality. Results from a Vancouver, BC, modeling study
(2003), “The most recent National Water Quality Inven-
suggest that if all of Vancouver’s existing buildings were
tory reports that runoff from urbanized areas is the
retrofitted with green roofs over the next 50 yr, the health
leading source of water quality impairments to surveyed
of the area watershed could be restored to natural hy-
estuaries and the third-largest source of impairments to
drologic conditions in terms of flood risk, aquatic habi-
surveyed lakes.”
tat, and water quality (Graham and Kim, 2003). This would
Establishing vegetation on rooftops, commonly re-
occur because green roofs have the ability to filter nu-
ferred to as green roofs, is an emerging strategy for miti-
merous contaminants from rainwater that has flowed
gating stormwater runoff (Monterusso et al., 2004; Moran
across the roof surface (Dramstad et al., 1996). Although
et al., 2003; Rowe et al., 2003; Schade, 2000). In addition,
minimal, Bucheli et al. (1998) detected concentrations
green roofs offer numerous other benefits beyond storm-
of three common classes of pesticides in non-green roof
water mitigation. They provide insulation for buildings,
runoff due to atmospheric deposits. Other studies showed
roof runoff contained higher amounts of numerous heavy
N.D. VanWoert, D.B. Rowe, and R.T. Fernandez, Department of
metals and nutrients when compared with rainfall, prob-
Horticulture; J.A. Andresen, Department of Geography; C.L. Rugh,
Department of Crop and Soil Sciences; and L. Xiao, College of Agri-
ably due to the runoff picking up particulate pollutants
culture and Natural Resources Statistical Consulting Center, Michigan
when flowing across the roof (Mason et al., 1999). For
State University, East Lansing, MI 48824. This paper is a portion of
green roofs, these pollutants can be taken up and de-
a thesis submitted by N.D. VanWoert. Received 27 Sept. 2004. Techni-
graded by the plants or bound in the growing substrate
cal Reports. *Corresponding author (rowed@msu.edu).
of green roofs (Johnston and Newton, 1996). Zobrist et al.
Published in J. Environ. Qual. 34:1036–1044 (2005).
(2000) concluded that without corrective measures, roof
doi:10.2134/jeq2004.0364
runoff pollutants will lower the water quality of surround-
© ASA, CSSA, SSSA
677 S. Segoe Rd., Madison, WI 53711 USA ing water bodies.
1036
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
VANWOERT ET AL.: GREEN ROOF STORMWATER RETENTION 1037
Fig. 1. Graphic representation of the model-scale roof platforms used to evaluate stormwater retention in the roof surface comparison study.
ville, Denver, CO), composed of a closed cell polyisocyanurate
An estimated 14% of all flat roofs are green in Ger-
foam core and fiberglass reinforced facers. Above the ENRGY
many, a nation widely considered the leader in green
2 layer was a 1.9-cm-thick insulation layer of Fesco board con-
roof research, technology, and usage (Herman, 2003). In
sisting of expanded perlite, blended with selected binders and
North America, the concept of green roofs is in its in-
fibers (Johns Manville). The top layer was a combination of
fancy. If green roof installations are to become common-
Paradiene 20 (Siplast, Irving, TX), a flexible membrane with an
place in the United States, quantifiable data that docu-
elastomeric asphalt base, and Teranap (Siplast), a polyester mat
ment the ability of green roofs to retain stormwater under
coated with styrene butadiene styrene (SBS)-modified bitumen,
the climatic conditions of the region must be available.
with a root-resistant polyester film covering the top side.
Data of this nature exist for particular drainage systems
Aluminum sheet metal troughs were attached on the low
end of the platforms to direct stormwater runoff through the
in other areas of the continent and Europe, but most
measuring devices used to quantify runoff. Each trough was
is not transferable to these specific climatic conditions.
divided into three separate sections corresponding to the three
Also, much of the current information is anecdotal in
divided sections. The wood-framed platforms included sides
nature, the information is proprietary, or the experi-
that extended 20.3 cm above the platform deck, also covered
ments were not performed in a replicated study. There-
with the waterproofin g membrane. Pla tforms were set at a 2%
fore, our objective was to quantify the differences in water
slope with the top edge of the high end 0.9 m above ground
retention among an extensive green roof, an extensive
level and oriented with the low end of the slope facing south
green roof without vegetation, and a standard gravel
to maximize sun exposure.
ballast roof in a replicated study. In addition, studies
Drainage System and Vegetation Carrier
were performed with the objective to quantify the differ-
ences in water retention among various substrate depths
Two of the three self-contained sections on each platform
and roof slopes in a replicated study. This information
used the Xero Flor XF108 drainage layer (Wolfgang Behrens
can then be used to make decisions concerning green roof
Systementwicklung GmbH, Groß Ippener, Germany) installed
over the Teranap Waterproofing System (Fig. 2). The drainage
usage to mitigate stormwater runoff and can potentially
be used to develop models to predict stormwater runoff
during the design of green roof systems.
MATERIALS AND METHODS
Study 1
Platforms
Three simulated roof platforms with overall dimensions of
2.44 ⫻ 2.44 m were constructed by ChristenDetroit (Detroit,
MI) at the Michigan State University Horticulture Teaching
and Research Center (East Lansing, MI) (Fig. 1). Each plat-
form simulated a commercial roof, including an insulation
layer, protective layers, and waterproofing membrane. How-
ever, since there was not an environmentally controlled room
under the platform, heat flux through the roof can be dis-
counted. Platforms were divided into three equal sections mea-
suring 0.67 ⫻ 2.44 m using wood dividers that were also cov-
ered with the waterproofing membrane. Lining the platform
Fig. 2. Cross-section of a representative extensive green roof system
including typically used layers.deck was 3.8 cm of ENRGY 2 insulation board (Johns Man-
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
1038 J. ENVIRON. QUAL., VOL. 34, MAY–JUNE 2005
Table 1. Potential water retention capacity of the green roof sys-
ment. Seedlings were acclimated from Days 52 through 57
tem components used in Studies 1 and 2.
by periodically removing the shade cloth depending on the
intensity of the sun, after which it was removed permanently.
Component Rainfall retention capacity
Upon seed distribution, an automated overhead irrigation
mm
system (Rainbird, Azusa, CA) was programmed to run six 10-min
Water retention fabric (1.5 cm) 2
cycles daily (0900, 1100, 1300, 1500, 1700, and 1900 h) through
Media (2.5 cm) 5
15 July 2002. From 16 July until 31 July 2002, the irrigation was
Media (4.0 cm) 8
Media (6.0 cm) 12
reduced to four 10-min cycles daily (0900, 1300, 1700, and 1900 h).
Irrigation was terminated on 31 July 2002 once the plants had
become established an d had achieved 100% coverage.
layer consisted of a geotextile fabric with nylon coils attached
on the underside. The total thickness of this layer was approx-
Roof Treatments
imately 1.5 cm. For additional water holding capacity, a 0.75-
cm-thick water retention fabric (Xero Flor XF158) capable of
Three roof types were tested: an extensive green roof with
retaining up to 1200 g m
⫺
2
of water was placed over the drain-
vegetation, an extensive green roof without vegetation (media-
age layer. The water retention fabric was composed of a recy-
only), and a conventional commercial roof with a 2-cm depth
cled synthetic fiber mixture consisting of polyester, polyamide,
gravel ballast. A gravel ballast is commonly used on flat com-
polypropylene, and acrylic fibers. Above this additional re-
mercial roofs to hold the waterproofing membrane in place.
tention fabric was the vegetation carrier (Xero Flor XF301),
The vegetated and media-only sections each contained a green
which included a recycled synthetic fiber fabric similar to
roof drainage system and vegetation carrier as described pre-
XF158 used for water retention sewn to an inverted layer of
viously. Roof treatments were arranged in a randomized com-
XF108 that held media and vegetation. This water retention
plete block design (RCBD) with three replications; each plat-
layer could hold up to 800 g m
⫺
2
of water and was approx-
form represented one block and the vegetation, media-only,
imately 0.75 cm thick. There was then 2.5 cm of growing media
or gravel ballast treatment was randomly assigned within sec-
placed on the vegetation carrier. The water retention fabric in
tions of each platform (Fig. 1).
combination with the 2.5 cm of growing media have the poten-
tial to hold up to 7 mm of rainfall (Table 1). Total thickness of
Data Collection and Analysis
the drainage layer, vegetation carrier, and growing media was
Model TE525WS tipping bucket rain gauges (Campbell Sci-
approximately 5.5 cm. The system as a whole permits water ex-
entific, Logan, UT) were mounted under the drain of each
ceeding the holding capacity of the retention fabric and plant-
platform section to quantify stormwater runoff. An additional
ing media to drain through the nylon coils and exit the roof.
tipping bucket was mounted above each gravel section to record
precipitation, catching and releasing quantified water onto the
Plant Establishment
top end of the gravel surface. A Model CM6 automated weather
station (Campbell Scientific) was installed on the researchOne hundred percent coverage (no visible growing media)
was achieved on the vegetated section before the initiation of site to record meteorological parameters. The weather station
included an ambient air temperature and relative humiditydata collection. Plant species used in this study included golden
carpet (Sedum acre L.), stonecrop (S. album L., S. kamtschati- probe covered by a six-plate gill radiation shield. The weather
station also included instruments to measure wind speed andcum ellacombianum Fisch., and S. pulchellum Michx.), stone
orpine (S. reflexum L.), and two-row stonecrop (S. spurium Bieb. direction as well as photosynthetically active radiation.
Data from the tipping bucket rain gauges and tripod weather‘Coccineum’ and ‘Summer Glory’). The plant mix was applied as
seed on 14 May 2002 at a rate of 1.3 g m
⫺
2
for each species . All station were collected at 5-min inter vals 24 h a day from 28
Aug. 2002 through 31 Oct. 2003 using a Campbell Scientificseeds were evenly mixed in dry sand to ensure even distribution
when the mixture was sown by hand on the platforms. Seeds CR10X datalogger equipped with switch closure modules and
a storage module. Accuracy of the rain gauges was ⫾1%, ⫹0were obtained from Jelitto Staudensamen GmbH (Schwarm-
stedt, Germany). and ⫺2.5%, and ⫹0 and ⫺3.5% for rainfalls of ⬍25.4 mm
h
⫺
1
, 25.4 to 50.8 mm h
⫺
1
, and 50.8 to 76.2 mm h
⫺
1
, respectively.Growing media consisted of 40% heat-expanded slate (gra-
dation 3–5 mm) (PermaTill; Carolina Stalite Company, Salis- During the largest rain event over the course of the study,
96.8% of the rain that fell on the conventional gravel roofbury, NC), 40% USGA (United States Golf Association)-
grade sand (Osburn Industries, Taylor, MI), 10% Michigan platforms was recorded exiting the roofs by the tipping bucket
rain gauges. The other 3.2% either evaporated or can be at-Peat (Osburn Industries), 5% dolomite (Osburn Industries),
3.33% composted yard waste (Kalamazoo Landscape Supplies, tributed to error. Although the raw data indicate otherwise,
runoff values from the conventional roof platforms with theKalamazoo, MI), and 1.67% composted poultry litter (Her-
bruck’s, Saranac, MI) by volume. Media bulk density, capillary gravel ballasts may be underestimated during some of the
heavier rain events.pore space, noncapillary pore space, infiltration rate, and
water holding capacity at 0.01 MPa were 130 kg m
⫺
3
, 19.9%, Retention data were analyzed from all rain events that oc-
curred d uring temperatures above 0⬚C as a percentage of total21.4%, 51.6 cm h
⫺
1
, and 17.1%, respectively (A & L Labora-
tories, Fort Wayne, IN). Saturated weight was equal to 150 kg rainfall for each rain event. Frozen precipitation was not physi-
cally removed from the platforms. Melting precipitation wasm
⫺
3
. At time of planting, electrical conductivity (EC) and pH
of the media were 0.33 S m
⫺
1
and 7.9, respectively. Each green allowed into the data set if it fully occurred in temperatures
above 0⬚C. Independent rain events were defined as precipi-roof system platform section was filled with planting media
to a depth of 2.5 cm. All sections of the platforms, except gravel, tation events separated by six or more hours. In the event
runoff was still occurring six hours after the first event, thehad 100 g m
⫺
2
of Nutricote Type 100, 20N–7P
2
O
5
–10K
2
O con-
trolled release fertilizer (Agrivert, Webster, TX) hand-applied two events were combined. Rain events were arbitrarily cate-
gorized as light (⬍2 mm), medium (2–6 mm), or heavy (⬎6 mm).at the time of planting and on 19 May 2003.
Platforms were covered with a plastic shade cloth (Wolfgang The extent of each category was chosen to get rain event
sample sizes that were similar across all three categories.Behrens Systementwicklung GmbH) for the first 52 d after
the seed was sown to enhance germination and plant establish- Data were analyzed as mean percent retention per rain
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
VANWOERT ET AL.: GREEN ROOF STORMWATER RETENTION 1039
event using an ANOVA model with platform as a random
effect and roof treatment and rainfall category as fixed effects.
Although original means are presented, all runoff values were
transformed before analysis using a power transformation (0.4)
to stabilize the variance and normalize the data. Significant
differences between treatments were determined using mul-
tiple comparisons with Tukey–Kramer adjustments (PROC
MIXED, SAS Version 8.02; SAS Institute, 2001). Total reten-
tion values for the study are presented, but were not subjected
to statistical analysis due to the limited number of data points.
Study 2
Twelve additional roof platforms were used to examine
roof slope and media depth. These platforms were constructed
as previously described, except that each 2.44- ⫻ 2.44-m plat-
form was considered an experimental unit as the platforms
were not divided into three equal sections. All platforms had
vegetated extensive green roof systems installed as described
previously and were subjected to the same environmental
conditions, and runoff data were collected with identical in-
Fig. 3. Daily precipitation (mm) during the experimental study (28
strumentation and protocols.
Aug. 2002 through 31 Oct. 2003). Values are averages of measure-
ments taken using three tipping bucket rain gauges mounted at
Treatments were arranged in a completely randomized de-
the research site.
sign (CRD) with three replications. Six platforms were set at
a 2% slope and six were set at a 6.5% slope. A total of three
tected from the gravel ballast treatment. The start of
growing media depths were examined, with two depths tested
at each slope. For the 2% slope, media depths of 2.5 and
runoff from the gravel treatments was delayed 10 min
4.0 cm were tested while depths of 4.0 and 6.0 cm were tested
past the initial rainfall during the representative medium
on the 6.5% slope platforms. Potential water retention capac-
rain event, and 15 min for both the media-only and
ity of the water retention fabric and growing media is shown
vegetated treatments. Following a delay of 15 min after
in Table 1.
the initial rainfall, runoff from all treatments was de-
Data were analyzed as mean percent retention per rain
event using an ANOVA model with roof slope, media depth,
and rainfall category as fixed effects. Although original means
are presented, all retention values were transformed before
analysis using a power transformation (0.113) to stabilize the
variance and normalize the data set. Significant differences
between treatments were determined using multiple compari-
sons with Tukey–Kramer adjustments (PROC MIXED, SAS
Version 8.02; SAS Institute, 2001). Total retention values for
the study are presented, but were not subjected to statistical
analysis due to the limited number of data points.
RESULTS
Measurable precipitation (⬎0 mm) was recorded on
162 of the 430 d of the study (38%) (Fig. 3). Daily pre-
cipitation amounts ranged from 0.08 to 53.59 mm. Of
the 83 rain events measured during temperatures above
0⬚C, there were 26 light (⬍2 mm), 30 medium (2–6 mm),
and 27 heavy (⬎6 mm) rain ev ents. Gener ally, low-v ol-
ume rain events were more frequent than larger rain
events. Daily maximum and minimum ambient air tem-
peratures ranged from ⫺9.9 to 34.2⬚C and ⫺24.6 to
20.8⬚C, respectively.
Study 1
Representative hydrographs (Fig. 4) and cumulative
hydrographs (Fig. 5) from a selected rain event within
each rainfall category show the effects that the roof treat-
Fig. 4. Runoff hydrographs of selected representative (A) light
(1.27 mm), (B) medium (4.06 mm), and (C) heavy (10.08 mm) rain
ments had on quantity , delay of the start, and time dura-
events recorded at 5-min intervals. Lines represent either rainfall
tion of runoff. During a represen tativ e light rain event,
(mm) or runoff (mm) from conventional roofs with a gravel ballast
the start of runoff from the vegetated treatments did not
(gravel), nonvegetated green roofs with media only (media), or
begin until 55 min after the initial rainfall was measured.
vegetated green roof treatments (vegetated). Values are averages
of measurements taken using three tipping bucket rain gauges.This delay was 1 5 min after the time when runoff was de-
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
1040 J. ENVIRON. QUAL., VOL. 34, MAY–JUNE 2005
Fig. 6. Retention percentage (%) averaged for all measured rain
events in respective categories (light, n ⫽ 26; medium, n ⫽ 30;
heavy, n ⫽ 27; overall, n ⫽ 83) for each roof treatment. Letters
above bars represent mean separation among treatments within
each rainfall category by Tukey’s Studentized Range (HSD) test,
P ⱕ 0.05, n ⫽ 3. Error bars represent standard error symmetrical
around the mean, but only the positive side is shown on the graph.
only and vegetated treatments each retained greater
than 96% of the rainfall. For combined medium rain
events (113 mm), the gravel ballast treatment retained
the least (33.9%) and the vegetated treatment retained
Fig. 5. Cumulative runoff hydrographs of selected representative (A)
the most (82.9%) rainfall. The same trend occurred for
light (1.27 mm), (B) medium (4.06 mm), and (C) heavy (10.08 mm)
rain events recorded at 5-min intervals. Lines represent either
combined heavy rain events (418 mm) with gravel bal-
rainfall (mm) or runoff (mm) from conventional roofs with a gravel
last retaining 22.2% and vegetated retaining 52.4% of
ballast (gravel), nonvegetated green roofs with media only (media),
the rainfall (Table 2).
or vegetated green roof treatments (vegetated). Values are aver-
When rainfall was separated into distinct rain events
ages of measurements taken using three tipping bucket rain gauges.
and retention percentages from each rain event were
averaged together, retention percentages were lowest
tected within 5 min of each other during the represen-
for the gravel ballast, followed by the media-only, and
tative heavy rain event. Runoff was not only delayed
vegetated roof treatments; all means were different (P ⱕ
during the heavy rain event with the media-only and
0.05) (Fig. 6). However, when the rain events were cate-
vegetated treatments, it was spread out over time; the
gorized into light, medium, and heavy, the media-only
last measured runoff was recorded nearly 3 h after the
and the vegetated treatments were not different in any
rain event ended, which was 30 min past the last runoff
of the rainfall categories, although both were different
from the gravel ballast treatment.
from the gravel ballast treatment. The lowest retention
Over the 14-mo period, the vegetated roof treatment
percentage for all treatments occurred during heavy rain
retained 337 mm of the 556 mm of cumulative rainfall
events where 26.3, 52.6, and 65.0% was retained for the
from the 83 measured rain events (60.6%) (Table 2).
gravel ballast, media-only, and vegetated treatments, re-
The media-only treatment retained 281 mm (50.4%)
spectively. During medium rain events, the media-only
and, as expected, the gravel ballast roof retained the
and vegetated treatments each retained an average of
least rainfall, 151 mm (27.2%). When total rainfall from
85.7% of the rainfall per rain event. The gravel ballast
all light rain events was combined (25 mm), the media-
treatment retained an average of 37.7% of the rainfall
for these events. The gravel ballast treatment retained
Table 2. Percentage of total rainfall retention over the 14-mo
an average of 84.6% of the rainfall for the light rain
period (28 Aug. 2002 to 31 Oct. 2003) from three roof platform
treatments replicated three times.
events, followed by the vegetated treatment (97.9%)
and media-only (99.6%).
Treatment† Light‡ Medium Heavy Overall
All treatments retained 100% of the rainfall from a
%
rain event on several occasions. This occurred seven, fif-
Gravel 79.9 33.9 22.2 27.2
teen, and twenty times on the grav el ballast, media-only ,
Media 99.3 82.3 38.9 50.4
Vegetated 96.2 82.9 52.4 60.6
and vegetated treatments, respectively. The heaviest rain-
fall for which 100% retention was achieved for the vege-
† Values denote retention from conventional roofs with a gravel ballast
(gravel), nonvegetated green roofs with media only (media), and vege-
tated treatment was 5.56 mm. This was likely possible
tated green roofs (vegetated).
because substrate moisture content was relatively low
‡ Rain event categories are light (⬍2 mm) (n ⫽ 26), medium (2–6 mm)
(n ⫽ 30), heavy (⬎6 mm) (n ⫽ 27), and overall (n ⫽ 83).
before the rain event. There was zero precipitation dur-
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
VANWOERT ET AL.: GREEN ROOF STORMWATER RETENTION 1041
Fig. 7. Runoff hydrographs of selected representative (A) light
Fig. 8. Cumulative runoff hydrographs of selected representative (A)
(1.27 mm), (B) medium (4.06 mm), and (C) heavy (10.08 mm) rain
light (1.27 mm), (B) medium (4.06 mm), and (C) heavy (10.08 mm)
events recorded at 5-min intervals. Lines represent either rainfall
rain events recorded at 5-min intervals. Lines represent either
(mm) or runoff (mm) from vegetated green roof platforms set at
rainfall (mm) or runoff (mm) from vegetated green roof platforms
a 2% roof slope with 2.5 cm of media (2%–2.5 cm), 2% roof slope
set at a 2% roof slope with 2.5 cm of media (2%–2.5 cm), 2% roof
with 4 cm of media (2%–4 cm), 6.5% roof slope with 4 cm of media
slope with 4 cm of media (2%–4 cm), 6.5% roof slope with 4 cm
(6.5%–4 cm), or 6.5% roof slope with 6 cm of media (6.5%–6 cm).
of media (6.5%–4 cm), or 6.5% roof slope with 6 cm of media
Values are averages of three replications measured using tipping
(6.5%–6 cm). Values are averages of three replications measured
bucket rain gauges mounted at the research site.
using tipping bucket rain gauges mounted at the research site.
ing the previous five days and the average high ambient
Individually, the 6.5% sloped platforms containing 4 cm
temperature was 29.8⬚C. For the gravel ballast treat-
of media retained the least amount of rainfall (65.9%)
ment, the heaviest event with complete retention was
(Table 3). Retention ranged from 97.1% (2%–2.5 cm)
0.76 mm. The least retention (12%) from the vegetated
during light rain events to 59.5% (6.5%–4 cm) for heavy
treatment occurred during a rain event of 73 mm that
rain events (Table 3).
spanned three days. Individual rain event retention per-
When total rainfall was separated into distinct rain
centages under 15% occurred numerous times for the
events and retention percentages were averaged together,
gravel ballast treatment during rain events ranging from
overall retention percentages ranged from 83.8% (6.5%–
0.68 to 73 mm.
4 cm) to more than 87% (2%–4 cm) when light, medium,
and heavy rain events were combined (Fig. 9). Overall,
Study 2
the greatest retention percentage (87%) occurred at
Representative hydrographs (Fig. 7) and cumulative
2%–4 cm.
hydrographs (Fig. 8) from a selected rain event within
When light, medium, and heavy rain events were cate-
each rainfall category show the effect that slope and
Table 3. Percentage of total rainfall retention over the 14-mo
media depth had on quantity of runoff, as well as their
period (28 Aug. 2002 to 31 Oct. 2003) from four roof platform
ability to delay runoff. Initial runoff from all four treat-
treatments replicated three times.
ments occurred within 10 min of each other for both
Treatment† Light‡ Medium Heavy Overall
the medium and heavy representative rain events. Dur-
%
ing the representative light rain event, runoff from both
2%–2.5 cm 95.1 82.9 64.7 69.8
treatments with 4 cm of media was delayed 30 to 40 min
2%–4.0 cm 97.1 85.5 65.1 70.7
compared with the 2%–2.5 cm and 6.5%–6 cm treat-
6.5%–4.0 cm 94.9 83.1 59.5 65.9
ments. Runoff was not only delayed during the repre-
6.5%–6.0 cm 95.8 84.6 62.0 68.1
sentative heavy rain event, it was spread out over time;
† Values denote retention from vegetated roof platforms set at a 2% roof
the last measured runoff from the platforms occurred
slope with 2.5 cm of media (2%–2.5 cm), 2% roof slope with 4 cm of
media (2%–4 cm), 6.5% roof slope with 4 cm of media (6.5%–4 cm),
14 h after the rain event ended.
or 6.5% roof slope with 6 cm of media (6.5%–6 cm).
Over the 14-mo period, the roof platforms retained
‡ Rain event categories are light (⬍2 mm) (n ⫽ 26), medium (2–6 mm)
(n ⫽ 30), heavy (⬎6 mm) (n ⫽ 27), and overall (n ⫽ 83).
more than 68% of the 556 mm of the measured rainfall.
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
1042 J. ENVIRON. QUAL., VOL. 34, MAY–JUNE 2005
which is very porous and allows for a higher water hold-
ing capacity relative to the open spaces within the gravel
ballast typically found on conventional roofs (Liesecke,
1998). The largest difference between the vegetated and
gravel ballast treatments occurred during medium rain
events when the vegetated treatment retained an aver-
age of 48% more w ater p er rain even t. The media-only
and vegetated treatments were not significantly differ-
ent when the rain events were categorized. This suggests
that the main factor for water retention is the physical
properties of the media as well as the presence of the
water retention fabric. In this e xperi ment, approximately
40% of the subst rate w as com posed of retention fabric.
The vegetated treatments retained 60% of the rainfall
they rece ived d uring the measured rain events, which is
about 10% higher than the finding s of Monteru sso et al.
(2004), but similar to the findings of Liesecke (1998)
and Schade (2000) when similarly designed green roof
Fig. 9. Retention percentage (%) for all measured rain events in re-
systems were used. The discrepancy between this study
spective categories (light, n ⫽ 26; medium, n ⫽ 30; heavy, n ⫽ 27;
and that of Monterusso et al. (2004) is probably due to
overall, n ⫽ 83) for each roof slope and media depth treatment.
Treatments were as follows: 2% roof slope with 2.5 cm of media
the lower number of rain events measured in the Monte-
(2%–2.5 cm), 2% roof slope with 4 cm of media (2%–4 cm), 6.5%
russo et al. (2004) study. Past studies have offered results
roof slope with 4 cm of media (6.5%–4 cm), and 6.5% roof slope
of retention per year percentages. However, they are
with 6 cm of media (6.5%–6 cm). Letters above bars represent
not possible with the data collected from this study be-
mean separation among treatments within each rainfall category
by Tukey’s Studentized Range (HSD) test, P ⱕ 0.05, n ⫽ 3. Error
cause the tipping bucket rain gauges did not function
bars represent standard error symmetrical around the mean, but
properly in temperatures below 0⬚C. However, we could
only the positive side is shown on the graph.
assume lower retention percentages during the winter
months in a locale such as Michigan, because evapotran-
gorized, the lowest retention percentage occurred dur-
spiration and soil infiltration are greatly reduced during
ing heavy rain events (69.2–75.6%). Light rain events
this time (Liesecke, 1998).
resulted in the highest retention percentage where more
Several studies have shown a delay in peak flow of
than 96% of the rainfall was retained regardless of roof
runoff from a green roof when compared with a standard
slope or media depth. The lowest retention percentage re-
roof (Liesecke, 1999; Moran et al., 2003; Schade, 2000).
corded during the study was 22%, which occurred dur-
From both plots during the representative heavy rainfall
ing a 2.37-mm rain event. One hundred percent reten-
event (Fig. 4 and 5), we can see that a delay in the onset
tion occurred on several occasions with rainfalls up to
of runoff on the green roof treatment is evident when
5.8 mm. The heaviest rain event, 73 mm, occurred on
compared with the gravel ballast. No delay can be seen
3–5 Apr. 2003. Thirty percent of the rain from this event
for the light and medium rainfall events due to the green
was retained when averaged over all four treatments.
roof treatments retaining nearly all of the rainfall. The
Retention percentages for light and medium rain
cumulative hydrographs offer another valuable method
events were greatest on the 2%–4 cm platforms (P ⱕ
of looking at the reduction green roofs provide. In all
0.05). The other three treatments were not different
three cumulative hydrographs, runoff from the gravel
from each other in these rainfall categories. For heavy
ballast treatment is evident unlike the representative
rain events, no difference was detected between treat-
plots for the media-only and vegetated treatments. The
ments. The 2%–4 cm treatment had the highest mean
peak flow reduction and the tendency to extend the
retention percentage when all rain events were com-
runoff over longer periods is very important for storm-
bined across rainfall categories. The other treatments
water management because the total amount of water
were again not significantly different from each other.
and rain event duration is often not the problem, it is
the rate that the incoming water needs to be treated.
Results of this study support earlier findings that green
DISCUSSION
roofs can reduce runoff from buildings. Past studies have
Study 1
indicated that media depth plays an important role in
water retention (Liesecke, 1998). From this information,
It was hypothesized that the gravel ballast roof would
we can imply that media and/or water retention fabric
yield considerably more runoff than the other two roof
is one of the most important factors for water retention.
treatments, but it was unclear if vegetation would be sig-
To our knowledge, the effect of vegetation relative to
nificantly different compared with the media-only treat-
media-only has not been studied even though it has gener-
ment. As expected, the gravel ballast roof retained less
ally been believed that vegetation plays a large role in
water in all rainfall categories when compared with the
water retention. However, our findings indicate that vege-
other two roof treatments on both a per rain event basis
tation is much less of an effect in aiding water retention
and for total rainfall. This occurrence is probably due to
the high surfac e area of the expanded slate-based m edia, when compared with media. Even so, vegetation plays
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
VANWOERT ET AL.: GREEN ROOF STORMWATER RETENTION 1043
other important roles such as preventing erosion of the will leave the roof as runoff regardless of media depth.
The only observed runoff delay among treatments oc-media from wind and water and providing transpirational
cooling and shade for the building, as well as mitigating curred during the representative light rain event. From
the results of this and previous studies, we can speculatethe urban heat island (Lu
¨
kenga and Wessels, 2001; Di-
moudi and Nikolopoulou, 2003). that roofs with deeper media provide a greater delay in
runoff due to increased water holding capacity.
Past studies have indicated that media moisture
Study 2
content immediately before a rain event influences the
Although Schade (2000) found similar runoff coeffi-
amount of water retained (Monterusso et al., 2004;
cients between four roof slopes using a vegetated mat
Moran et al., 2003). Rainfall intensity and duration also
green roof system, we hypothesized that increasing roof
play a part in water retention. Media moisture content,
slope would increase the quantity of runoff and that this
rainfall intensity, and rain event duration likely explain
occurrence could be offset by increased media depth.
differences between this study and others.
As expected, platforms built on a 2% slope containing
4 cm of media retained a greater quantity of rain than
CONCLUSIONS
the others on both a per rain event basis and for total
rainfall. Retention percentage for this treatment was
Vegetated platforms retained greater quantities of
significantly greater than the others in all rainfall cate-
stormwater than the conventional roofs with a gravel
gories except heavy events. Although the difference was
ballast. While vegetation did affect stormwater reten-
significant, the difference from other treatments was
tion, it was minimal relative to the effects of growing
minimal. No treatment consistently yielded the lowest
media. Media depth also influenced water retention on
retention value in all rainfall categories.
our model-scale extensive green roofs at one of the
Overall, at the 4-cm depth the treatments on a 2%
tested slopes. Other studies that have considered the
slope retained significantly more water than the 6.5%
effects of media depth on water retention have found
slope treatments. This finding contradicts those of earlier
similar results. However, our finding that retention per-
studies. Schade (2000) reported nearly constant water re-
centages were affected by the two slopes with equal
tention rates for roof slopes ranging from 2% up to
media depths contradicts results regarding roof slope
58%. Liesecke (1999) generalized that annual retention
reported by Liesecke (1999) and Schade (2000). If the
rates of 55 to 65% on an 8.7% sloped roof are compara-
objective of a green roof is to maximize rainfall reten-
ble to a 2% slope. The difference in findings between
tion, then factors such as slope and media depth must
past and current studies could be due to differences in
be addressed.
media composition among the studies.
Although green roofs are not new to other parts of the
Increasing media depth increased water retention at
world, they are a promising new technology to mitigate
only one slope. Retention percentages for platforms with
stormwater runoff quantity and quality in the United
6 cm of media were not different from platforms with
States. They are a technology that should be considered
4 cm of media on the 6.5% roof slope. However, for
for all roofing projects, especially those projects in areas
the 2% roof slope, deeper media (4 cm) retained a sig-
where stormwater management is a concern for city plan-
nificantly greater percentage of water for both the light
ners. With the continual increase of area covered by
and medium rainfall categories, but not heavy (P ⱕ
impervious surfaces, the already important problem of
0.05). Together with past studies, we can establish that
stormwater management will only become more of an
increasing media depth usually increases retention (Lie-
issue. Green roofs offer a new tool that shows promise
secke, 1998).
as a technology that can aid in providing a sustainably
Media depth should be considered for reasons other
built environment.
than just stormwater retention. In Quebec, Boivin et al.
(2001) found that substrate depth can influence freezing
ACKNOWLEDGMENTS
injury in certain herbaceous perennials. The researchers
Funding for this study was provided by Ford Motor Com-
concluded that in their climatic region, a minimum sub-
pany (Dearborn, MI), ChristenDETROIT Roofing Con-
strate depth of 10 cm should be used for the green roof
tractors (Detroit, MI), Wolfgang Behrens Systementwicklung
system constructed for their study. Other studies found
GmbH (Groß Ippener, Germany), and the Michigan Agri-
that media depth influences the growth, drought stress,
cultural Experiment Station. Illustrations drawn by Marlene
and drought tolerance of green roof vegetation (Durh-
Cameron.
man et al., 2004; Lassalle, 1998; Monterusso et al., 2005;
VanWoert et al., 2005).
REFERENCES
As mentioned previously, several studies have shown
Boivin, M., M. Lamy, A. Gosselin, and B. Dansereau. 2001. Effect
a delay in peak flow of runoff from a green roof when
of artificial substrate depth on freezing injury of six herbaceous
compared with a standard roof (Liesecke, 1999; Moran
perennials grown in a green roof system. HortTechnology 11:409–412.
Brenneisen, S. 2003. The benefits of biodiversity from green roofs—
et al., 2003; Schade, 2000). However, Fig. 8 shows that
Key design consequences. p. 323–329. In Proc. of 1st North Ameri-
the effect of roof slope and media depth on runoff delay
can Green Roof Conf.: Greening Rooftops for Sustainable Com-
is minimal for rain events greater than 2 mm. This im-
munities, Chicago. 29–30 May 2003. The Cardinal Group, Toronto.
plies that once sufficient rainfall has occurred to reach
Bucheli, T.D., S.R. Mu
¨
ller, S. Heberle, and R.P. Schwarzenbach. 1998.
Occurrence and behavior of pesticides in rainwater, roof runoff,the media’s water holding capacity, additional rainfall
Reproduced from Journal of Environmental Quality. Published by ASA, CSSA, and SSSA. All copyrights reserved.
1044 J. ENVIRON. QUAL., VOL. 34, MAY–JUNE 2005
and artificial stormwater infiltration. Environ. Sci. Technol. 32: Green Roof Conf.: Greening Rooftops for Sustainable Communi-
ties, Chicago. 29–30 May 2003. The Cardinal Group, Toronto.
3457–3464.
Lu
¨
kenga, W., and K. Wessels. 2001. Oberfla
¨
chentemperaturen von
Dimoudi, A., and M. Nikolopoulou. 2003. Vegetation in the urban
dachfla
¨
chen. Stadt Gru
¨
n 50:339–403.
environment: Microclimatic analysis and benefits. Energy Build.
Mason, Y., A.A. Ammann, A. Ulrich, and L. Sigg. 1999. Behavior of
35:69–76.
heavy metals, nutrients, and major components during roof runoff
Dramstad, W.E., J.D. Olson, and R.T.T. Forman. 1996. Landscape
infiltration. Environ. Sci. Technol. 33:1588–1597.
ecology. Principles in landscape architecture and land-use planning.
Monterusso, M.A., D.B. Rowe, and C.L. Rugh. 2005. Establishment
Harvard Univ. Graduate School of Design, Island Press, and Am.
and persistence of Sedum spp. and native taxa for green roof appli-
Soc. of Landscape Architects, Washington, DC.
cations. HortScience 40:391–396.
Durhman, A., N.D. VanWoert, D.B. Rowe, C.L. Rugh, and D. Ebert-
Monterusso, M.A., D.B. Rowe, C.L. Rugh, and D.K. Russell. 2004.
May. 2004. Evaluation of Crassulacean species on extensive green
Runoff water quantity and quality from green roof systems. Acta
roofs. p. 504–517. In Proc. of the 2nd North American Green Roof
Hortic. 639:369–376.
Conf.: Greening Rooftops for Sustainable Communities, Portland,
Moran, A., B. Hunt, and G. Jennings. 2003. A North Carolina field
OR. 2–4 June 2004. The Cardinal Group, Toronto.
study to evaluate greenroof runoff quality, runoff quantity, and plant
Gedge, D. 2003. From rubble to redstarts. p. 233–241. In Proc. of
growth. ASAE Paper 032303. Am. Soc. of Agric. Eng., St. Joseph, MI.
1st North American Green Roof Conf.: Greening Rooftops for
Niachou, A., K. Papakonstantinou, M. Santamouris, A. Tsangras-
Sustainable Communities, Chicago. 29–30 May 2003. The Cardinal
soulis, and G. Mihalakakou. 2001. Analysis of the green roof ther-
Group, Toronto.
mal properties and investigation of its energy performance. Energy
Giesel, D. 2001. Gru
¨
n auf das dach-kosten in den keller? Stadt Gru
¨
n
Build. 33:719–729.
50:404–406.
Rosenfeld, A., H. Akbariand, J. Romm, and M. Pomerantz. 1998.
Graham, P., and M. Kim. 2003. Evaluating the stormwater manage-
Cool communities: Strategies for heat island mitigation and smog
ment benefits of green roofs through water balance modeling.
reduction. Energy Build. 28:51–62.
p. 390–398. In Proc. of 1st North American Green Roof Conf.:
Rowe, D.B., C.L. Rugh, N. VanWoert, M.A. Monterusso, and D.K.
Greening Rooftops for Sustainable Communities, Chicago. 29–30
Russell. 2003. Green roof slope, substrate depth, and vegetation
May 2003. The Cardinal Group, Toronto.
influence runoff. p. 354–362. In Proc. of 1st North American Green
Herman, R. 2003. Green roofs in Germany: Yesterday, today and
Roof Conf.: Greening Rooftops for Sustainable Communities, Chi-
tomorrow. p. 41–45. In Proc. of 1st North American Green Roof
cago. 29–30 May 2003. The Cardinal Group, Toronto.
Conf.: Greening Rooftops for Sustainable Communities, Chicago.
SAS Institute. 2001. SAS Version 8.02. SAS Inst., Cary, NC.
29–30 May 2003. The Cardinal Group, Toronto.
Schade, C. 2000. Wasserru
¨
ckhaltung und Abflußbeiwerte bei du
¨
nn-
Johnston, J., and J. Newton. 1996. Building green. A guide for using
schichtigen extensivbegru
¨
nungen. Stadt Gru
¨
n 49:95–100.
plants on roofs, walls and pavements. The London Ecol. Unit, London.
USEPA. 2003. Protecting water quality from urban runoff. EPA 841-
Lassalle, F. 1998. Wirkung von trockenstreß auf xerophile pflanzen.
F-03-003. USEPA, Washington, DC.
Stadt Gru
¨
n 47:437–443.
VanWoert, N.D., D.B. Rowe, J.A. Andresen, C.L. Rugh, and L. Xiao.
Liesecke, H.J. 1998. Das retentionsvermo
¨
gen von dachbegru
¨
nungen.
2005. Watering regime and green roof substrate design impact
Stadt Gru
¨
n 47:46–53.
Sedum plant growth. HortScience 40 (in press).
Liesecke, H.J. 1999. Extensive begru
¨
nung bei 5⬚ dachneigung. Stadt
Wong, N.H., Y. Chen, C.L. Ong, and A. Sia. 2003. Investigation of
Gru
¨
n 48:337–346.
thermal benefits of rooftop garden in the tropical environment.
Liesecke, H.J., and H. Borgwardt. 1997. Abbau von luftschadstoffen
Build. Environ. 38:261–270.
durch extensive dachbegru
¨
nungen. Stadt Gru
¨
n 33:245–251.
Zobrist, J., S.R. Mu
¨
ller, A. Ammann, T.D. Bucheli, V. Mottier, M.
Liptan, T. 2003. Planning, zoning and financial incentives for ecoroofs
Ochs, R. Schoenenberger, J. Eugster, and M. Boller. 2000. Quality
of roof runoff for groundwater infiltration. Water Res. 34:1455–1462.in Portland, Oregon. p. 113–120. In Proc. of 1st North American
